Chapter 11 Introduction to Programming in C

Transcription

1 Programming Levels Chapter 11 Introduction to Programming in C Based on slides McGraw-Hill Additional material 2004/2005/2006 Lewis/Martin High-Level Languages Low-Level Languages Application Languages (Java, C#) Scripting Languages (Perl, Python, VB) System Programming Languages (C, C++) Assembly Language (x86, PowerPC, SPARC, MIPS) Machine Language (x86, PowerPC, SPARC, MIPS) Hardware (Application-Specific Integrated Circuits or ASICs) Interpreted Or Compiled Compilation Assembly 2 The Course Thus Far We did digital logic Bits are bits Ultimately, to understand a simple processor Why High-Level Languages? Easier than assembly. Why? Less primitive constructs Variables Type checking We did assembly language programming Programming the raw metal of the computer Ultimately, to understand C programming Portability Write program once, run it on the LC-3 or Intel s x86 Starting today: we re doing C programming C is still common for systems programming You ll need it for the operating systems class (CSE380) Ultimately, for a deeper understanding of any language (Java) Disadvantages Slower and larger programs (in most cases) Can t manipulate low-level hardware!all operating systems have some assembly in them Verdict: assembly coding is rare today 3 4

2 Our Challenge All of you already know Java We re going to try to cover the basics quickly We ll spend more time on poers & other C-specific nastiness Created two decades apart C: 1970s - AT&T Bell Labs C++: 1980s - AT&T Bell Labs Java: 1990s - Sun Microsystems Java and C/C++ Syntactically similar (Java uses C syntax) C lacks many of Java s features Subtly different semantics 5 C is Similar To Java Without: Objects No classes, objects, methods, or inheritance Exceptions Check all error codes explicitly Standard class library C has only a small standard library Garbage collection C requires explicit memory allocate and free Safety Java has strong type checking, checks array bounds In C, anything goes Portability Source: C code is less portable (but better than assembly) Binary: C compiles to specific machine code 6 More C vs Java differences C has a preprocessor A separate pre-pass over the code Performs replacements Include vs Import Java has import java.io.*; C has: #include <stdio.h> #include is part of the preprocessor Boolean type Java has an explicit boolean type C just uses an as zero or non-zero C s lack of boolean causes all sorts of trouble More differences as we go along 7 History of C and Unix Unix is the most influential operating system First developed in 1969 at AT&T Bell Labs By Ken Thompson and Dennis Ritchie Designed for smaller computers of the day Reject some of the complexity of MIT s Multics They found writing in assembly tedious Dennis Ritchie invented the C language in 1973 Based on BCPL and B, needed to be efficient (24KB of memory) Unix roduced to UC-Berkeley (Cal) in 1974 Bill Joy was an early Unix hacker as a PhD student at Cal Much of the early ernet consisted of Unix systems Mid-80s Good, solid TCP/IP for BSD in 1984 Linux - Free (re)implementation of Unix (libre and gratuit) Announced by Linus Torvalds in 1991 Much more in CSE380! 8

3 Aside: The Unix Command Line Text-based approach to give commands Commonly used before graphical displays Many advantages even today Examples mkdir cse240hw8 make a directory cd cse240hw8 change to the directory ls list contents of directory cp /mnt/eniac/home1/c/cse240/project/hw/hw8/*.! Copy files from one location to current dir (. ) emacs foo.c & run the command emacs with input foo.c gcc -o foo foo.c compile foo.c (create program called foo ) Unix eventually developed graphical UIs (GUIs) X-windows (long before Microsoft Windows) 9 What is C++? C++ is an extension of C Also done at AT&T Bell Labs (1983) Backward compatible (good and bad) That is, all C programs are legal C++ programs C++ adds many features to C Classes, objects, inheritance Templates for polymorphism A large, cumbersome class library (using templates) Exceptions (not actually implemented for a long time) More safety (though still unsafe) Operator and function overloading Kitchen sink Thus, many people uses it (to some extent) However, we re focusing on only C, not C++ 10 Quotes on C/C++ vs Java C is to assembly language as Java is to C Unknown "With all due respect, saying Java is just a C++ subset is rather like saying that `Pride and Prejudice' is just a subset of the Encyclopedia Britanica. While it is true that one is shorter than the other, and that both have the same syntax, there are rather overwhelming differences. Sam Weber, on the ACM SIGSCE mailing list Java is C++ done right. Unknown More quotes on C/C++ "The C programming language combines the power of assembly language with the ease-of-use of assembly language. Unknown "It is my impression that it's possible to write good programs in C++, but nobody does. John Levine, moderator of comp.compilers C makes it easy to shoot yourself in the foot; C++ makes it harder, but when you do it, it blows your whole leg off. Bjarne Stroustrup, creator of C

4 Program Execution: Compilation vs Interpretation Different ways of executing high-level languages Interpretation Interpreter: program that executes program statements!directly erprets program (portable but slow)!limited optimization Easy to debug, make changes, view ermediate results Languages: BASIC, LISP, Perl, Python, Matlab Compilation Compiler: translates statements o machine language!creates executable program (non-portable, but fast)!performs optimization over multiple statements Harder to debug, change requires recompilation Languages: C, C++, Fortran, Pascal Hybrid Java, has features of both erpreted and compiled languages 13 Compilation vs. Interpretation Consider the following algorithm: Get W from the keyboard. X = W + W Y = X + X Z = Y + Y Pr Z to screen. If erpreting, how many arithmetic operations occur? If compiling, we can analyze the entire program and possibly reduce the number of operations. Can we simplify the above algorithm to use a single arithmetic operation? 14 Compiling a C Program Entire mechanism is usually called the compiler Preprocessor Macro substitution Conditional compilation Source-level transformations!output is still C Compiler Generates object file!machine instructions Linker Combine object files (including libraries) o executable image Library Object Files Source Code Analysis Target Code Synthesis C Source and Header Files C Preprocessor Compiler Linker Executable Image Symbol Table 15 Compiler Source Code Analysis Front end Parses programs to identify its pieces!variables, expressions, statements, functions, etc. Depends on language (not on target machine) Code Generation Back end Generates machine code from analyzed source May optimize machine code to make it run more efficiently Very dependent on target machine Example Compiler: GCC The Free-Software Foundation s compiler Many front ends: C, C++, Fortran, Java Many back ends: Intel x86, PowerPC, SPARC, MIPS, Itanium 16

5 A Simple C Program #include <stdio.h> #define STOP 0 void main() /* variable declarations */ counter; /* an eger to hold count values */ startpo; /* starting po for countdown */ Preprocessor Directives #include <stdio.h> Before compiling, copy contents of header file (stdio.h) o source code. Header files typically contain descriptions of functions and variables needed by the program.!no restrictions -- could be any C source code /* prompt user for input */ prf("enter a positive number: "); scanf("%d", &startpo); /* read o startpo */ /* count down and pr count */ for (counter=startpo; counter >= STOP; counter--) prf("%d\n", counter); #define STOP 0 Before compiling, replace all instances of the string "STOP" with the string "0" Called a macro Used for values that won't change during execution, but might change if the program is reused. (Must recompile.) Comments main Function Begins with /* and ends with */ Every C program must have a function called main() Can span multiple lines Comments are not recognized within a string!example: "my/*don't pr this*/string" would be pred as: my/*don't pr this*/string Starting po for every program Similar to Java s main method!public static void main(string[] args) Begins with // and ends with end of line Single-line comment Much like ; in LC-3 assembly Introduced in C++, later back-ported to C The code for the function lives within brackets: void main() /* code goes here */ As before, use comments to help reader, not to confuse or to restate the obvious 19 20

6 Variable Declarations Input and Output Variables are used as names for data items Variety of I/O functions in C Standard Library Must include <stdio.h> to use them Each variable has a type, tells the compiler: How the data is to be erpreted How much space it needs, etc. counter; startpo; C has similar primitive types as Java, char, long, float, double More later prf("%d\n", counter); String contains characters to pr and formatting directions for variables This call says to pr the variable counter as a decimal eger, followed by a linefeed (\n) scanf("%d", &startpo); String contains formatting directions for looking at input This call says to read a decimal eger and assign it to the variable startpo (Don't worry about the & yet) More About Output Can pr arbitrary expressions, not just variables prf("%d\n", startpo - counter); Pr multiple expressions with a single statement prf("%d %d\n", counter, startpo - counter); Different formatting options: %d decimal eger %x hexadecimal eger %c ASCII character %f floating-po number Examples This code: prf("%d is a prime number.\n", 43); prf("43 plus 59 in decimal is %d.\n", 43+59); prf("43 plus 59 in hex is %x.\n", 43+59); prf("43 plus 59 as a character is %c.\n", 43+59); produces this output: 43 is a prime number. 43 plus 59 in decimal is plus 59 in hex is plus 59 as a character is f

7 Examples of Input Compiling and Linking Many of the same formatting characters are available for user input Various compilers available cc, gcc includes preprocessor, compiler, and linker scanf("%c", &nextchar); reads a single character and stores it in nextchar scanf("%f", &radius); reads a floating po number and stores it in radius scanf("%d %d", &length, &width); reads two decimal egers (separated by whitespace), stores the first one in length and the second in width Lots and lots of options! level of optimization, debugging preprocessor, linker options ermediate files -- object (.o), assembler (.s), preprocessor (.i), etc. Must use ampersand (&) for variables being modified (Explained in Chapter 16.) Remaining Chapters A more detailed look at many C features Variables and declarations Operators Control Structures Functions Poers and Data Structures I/O Emphasis on how C is converted to assembly language Chapter 12 Variables and Operators Also see C Reference in Appendix D Based on slides McGraw-Hill Additional material 2004/2005/2006 Lewis/Martin 27

8 Basic C Elements Variables Named, typed data items Operators Predefined actions performed on data items Combined with variables to form expressions, statements Statements and Functions Group together operations 29 Data Types C has several basic data types long float double char eger (at least 16 bits, commonly 32 bits) eger (at least 32 bits) floating po (at least 32 bits) floating po (commonly 64 bits) character (at least 8 bits) Exact size can vary, depending on processor is supposed to be "natural" eger size; for LC-3, that's 16 bits bits for most modern processors Signed vs unsigned: Default is 2 s complement signed egers Use unsigned keyword for unsigned numbers 30 Variable Names Any combination of letters, numbers, and underscore (_) Case sensitive "sum" is different than "Sum" Cannot begin with a number Usually, variables beginning with underscore are used only in special library routines Only first 31 characters are definitely used Implementations can consider more characters if they like Examples Legal i wordspersecond same identifier words_per_second _green areally_longname_morethan31chars areally_longname_morethan31characters Illegal 10sdigit ten'sdigit done? double reserved keyword 31 32

9 Literals Integer 123 /* decimal */ x123 /* hexadecimal */ Floating po e23 /* x */ 5E12 /* 5.0 x */ Character 'c' '\n' /* newline */ '\xa' /* ASCII 10 (0xA) */ 33 Scope: Global and Local Where is the variable accessible? Global: accessed anywhere in program Local: only accessible in a particular region Compiler infers scope from where variable is declared Programmer doesn't have to explicitly state Variable is local to the block in which it is declared Block defined by open and closed braces Can access variable declared in any "containing" block Global variable is declared outside all blocks 34 Example #include <stdio.h> itsglobal = 0; main() itslocal = 1; /* local to main */ prf("global %d Local %d\n", itsglobal, itslocal); itslocal = 2; /* local to this block */ itsglobal = 4; /* change global variable */ prf("global %d Local %d\n", itsglobal, itslocal); prf("global %d Local %d\n", itsglobal, itslocal); Output Global 0 Local 1 Global 4 Local 2 Global 4 Local 1 Expression Any combination of variables, constants, operators, and function calls Every expression has a type, derived from the types of its components (according to C typing rules) Examples: counter >= STOP x + sqrt(y) x & z w-- %

10 Statement Expresses a complete unit of work Executed in sequential order Simple statement ends with semicolon z = x * y; /* assign product to z */ y = y + 1; /* after multiplication */ ; /* null statement */ Compound statement formed with braces Syntactically equivalent to a simple statement z = x * y; y = y + 1; Operators Three things to know about each operator (1) Function What does it do? (2) Precedence In which order are operators combined? Example: "a * b + c * d" is the same as "(a * b) + (c * d)" because multiply (*) has a higher precedence than addition (+) (3) Associativity In which order are operators of the same precedence combined? Example: "a - b - c" is the same as "(a - b) - c" because add/sub associate left-to-right Assignment Operator Assignment Operator Changes the value of a variable All expressions evaluate to a value,even ones with the assignment operator x = x + 4; 1. Evaluate right-hand side. 2. Set value of left-hand side variable to result. For assignment, the result is the value assigned Usually (but not always) the value of the right-hand side!type conversion might make assigned value different than computed value Assignment associates right to left. y = x = 3; y gets the value 3, because (x = 3) evaluates to the value 3 y = (x = 3); 39 40

11 Arithmetic Operators Symbol Operation Usage Precedence Assoc * multiply x * y 6 l-to-r / divide x / y 6 l-to-r % modulo x % y 6 l-to-r + addition x + y 7 l-to-r - subtraction x - y 7 l-to-r All associate left to right * / % have higher precedence than + - Example * 4 (2 + 3) * 4 versus Arithmetic Expressions If mixed types, smaller type is "promoted" to larger x if x is, converted to double and result is double Integer division -- fraction is dropped x / 3 if x is and x=5, result is 1 (not ) Modulo -- result is remainder x % 3 if x is and x=5, result is Bitwise Operators Symbol Operation Usage Precedence Assoc ~ bitwise NOT ~x 4 r-to-l << left shift x << y 8 l-to-r >> right shift x >> y 8 l-to-r & bitwise AND x & y 11 l-to-r ^ bitwise XOR x ^ y 12 l-to-r bitwise OR x y 13 l-to-r Operate on variables bit-by-bit Like LC-3 AND and NOT instructions Shift operations are logical (not arithmetic) Operate on values -- neither operand is changed x = y << 1 same as x = y+y Logical Operators Symbol Operation Usage Precedence Assoc! logical NOT!x 4 r-to-l && logical AND x && y 14 l-to-r logical OR x y 15 l-to-r Treats entire variable (or value) as TRUE (non-zero), or FALSE (zero). Result is 1 (TRUE) or 0 (FALSE) x = 15; y = 0; prf( %d, x y); Bit-wise vs Logical 1 & 8 = 0 ( AND = ) 1 && 8 = 1 (True & True = True) 43 44

12 Relational Operators Symbol Operation Usage Precedence Assoc > greater than x > y 9 l-to-r >= greater than or equal x >= y 9 l-to-r < less than x < y 9 l-to-r <= less than or equal x <= y 9 l-to-r == equal x == y 10 l-to-r!= not equal x!= y 10 l-to-r Result is 1 (TRUE) or 0 (FALSE) Assignment vs Equality Don't confuse equality (==) with assignment (=) x = 9; y = 10; if (x == y) prf( not executed\n ); if (x = y) prf( %d %d, x, y); Result: is pred. Why? Compiler will not stop you! (What happens in Java?) Special Operators: ++ and -- Changes value of variable before (or after) its value is used in an expression Symbol Operation Usage Precedence Assoc ++ postincrement x++ 2 r-to-l -- postdecrement x-- 2 r-to-l ++ preincrement ++x 3 r-to-l -- predecrement --x 3 r-to-l Pre: Increment/decrement variable before using its value Post: Increment/decrement variable after using its value Using ++ and -- x = 4; y = x++; Results: x = 5, y = 4 (because x is incremented after assignment) x = 4; y = ++x; Results: x = 5, y = 5 (because x is incremented before assignment) Please, don t combine ++ and =. Really. Just don t! 47 48

13 Special Operators: +=, *=, etc. Arithmetic and bitwise operators can be combined with assignment operator Statement Equivalent assignment x += y; x = x + y; x -= y; x = x - y; x *= y; x = x * y; x /= y; x = x / y; x %= y; x = x % y; All have same x &= y; x = x & y; precedence and associativity as = x = y; x = x y; and associate x ^= y; x = x ^ y; right-to-left. x <<= y; x = x << y; x >>= y; x = x >> y; 49 Special Operator: Conditional Symbol Operation Usage Precedence Assoc?: conditional x?y:z 16 l-to-r x? y : z If x is non-zero, result is y If x is zero, result is z Seems useful, but I don t use it A normal if is almost always more clear You don t need to use every language feature Really, don t use it (you don t have to show how clever you are) 50 Practice with Precedence Assume a=1, b=2, c=3, d=4 x = a * b + c * d / 2; /* x = 8 */ same as: x = (a * b) + ((c * d) / 2); For long or confusing expressions, use parentheses, because reader might not have memorized precedence table Note: Assignment operator has lowest precedence, so all the arithmetic operations on the right-hand side are evaluated first 51 Practice with Operators In preparation for our dis-assembler (HW#8): opcode( ir) /* code to extract bits 15 through 12 of ir */ get_field( bits, hi_bit, lo_bit) /* code to extract hi_bit through lo_bit of bits */ For example, body of opcode function is now just get_field(ir, 15, 12); What about a signed-extended version? hi lo

14 Practice with Operators (Solution 1) Practice with Operators (Solution 2) opcode( ir) ir = ir >> 12; ir = ir & 0xf; return ir; OR opcode( ir) ir = ir & 0xf000; ir = ir >> 12; return ir; get_field( bits, hi_bit, lo_bit) inv_mask = ~0 << (hi_bit+1) mask = ~inv_mask; bits = bits & mask; // Mask off high-order bits bits = bits >> lo_bit; // Shift away low-order bits return bits; OR get_field( bits, hi_bit, lo_bit) bits = ~(~0 << (hi_bit+1)) & bits; // Mask high bits bits = bits >> lo_bit; // Shift away low-order bits return bits; Sign Extended Version get_sext_field( bits, hi_bit, lo_bit) most_significant_bit = bits & (1 << hi_bit); if (most_significant_bit!= 0) bits = (~0 << hi_bit) bits; // One extend else bits = ~(~0 << (hi_bit+1)) & bits; // Zero extend bits = bits >> lo_bit; // Shift away low-order bits return bits; Chapter 13 Control Structures Based on slides McGraw-Hill Additional material 2004/2005/2006 Lewis/Martin 55

15 Control Structures Conditional Making decision about which code to execute, based on evaluated expression if if-else switch If if (condition) action; condition T action F Iteration Executing code multiple times, ending based on evaluated expression while for do-while Condition is a C expression, which evaluates to TRUE (non-zero) or FALSE (zero). Action is a C statement, which may be simple or compound (a block) Example If Statements if (x <= 10) y = x * x + 5; if (x <= 10) y = x * x + 5; z = (2 * y) / 3; if (x <= 10) y = x * x + 5; z = (2 * y) / 3; Style: avoid singleton if statements (I really dislike them) compound statement; both executed if x <= 10 only first statement is conditional; second statement is always executed More If Examples if (0 <= age && age <= 11) kids = kids + 1; if (month == 4 month == 6 month == 9 month == 11) prf( The month has 30 days.\n ); if (x = 2) y = 5; Common C error, assignment (=) Versus equality (==) This is a common programming error (= instead of ==), not caught by compiler because it s syntactically correct

16 Generating Code for If Statement if (x == 2) y = 5; LDR R0, R6, #0 ; load x o R0 ADD R0, R0, #-2 ; subtract 2 BRnp NOT_TRUE ; if non-zero, x is not 2 AND R1, R1, #0 ; store 5 to y ADD R1, R1, #5 STR R1, R6, #1 NOT_TRUE... ; next statement If-else if (condition) action_if; else action_else; T action_if condition F action_else Else allows choice between two mutually exclusive actions without re-testing condition Generating Code for If-Else if (x) y++; z--; else y--; z++; LDR R0, R6, #0 BRz ELSE ; x is not zero LDR R1, R6, #1 ; incr y ADD R1, R1, #1 STR R1, R6, #1 LDR R1, R6, #2 ; decr z ADD R1, R1, #-1 STR R1, R6, #2 BR DONE ; skip else code Matching Else with If Else is always associated with closest unassociated if if (x!= 10) if (y > 3) z = z / 2; else z = z * 2; is NOT the same as... is the same as... if (x!= 10) if (y > 3) z = z / 2; else z = z * 2; ; x is zero ELSE LDR R1, R6, #1 ; decr y if (x!= 10) ADD R1, R1, #-1 if (y > 3) STR R1, R6, #1 z = z / 2; Solution: always use braces LDR R1, R6, #2 ; incr z ADD R1, R1, #1 (avoids the problem entirely) STR R1, R6, #2 else DONE... ; next statement 63 z = z * 2; 64

17 Chaining If s and Else s if (month == 4 month == 6 month == 9 month == 11) prf( Month has 30 days.\n ); else if (month == 1 month == 3 month == 5 month == 7 month == 8 month == 10 month == 12) prf( Month has 31 days.\n ); else if (month == 2) prf( Month has 28 or 29 days.\n ); else prf( Don t know that month.\n ); While while (test) loop_body; Executes loop body as long as test evaluates to TRUE (non-zero) test loop_body Note: Test is evaluated before executing loop body T F Generating Code for While x = 0; while (x < 10) prf( %d, x); x = x + 1; LOOP DONE AND R0, R0, #0 STR R0, R6, #0 ; x = 0 ; test LDR R0, R6, #0 ; load x ADD R0, R0, #-10 BRzp DONE ; loop body LDR R0, R6, #0 ; load x... <prf>... ADD R0, R0, #1 ; incr x STR R0, R6, #0 BR LOOP ; test again ; next statement Infinite Loops The following loop will never terminate: x = 0; while (x < 10) prf( %d, x); Loop body does not change condition......so test is never false Common programming error that can be difficult to find 67 68

18 For for (init; end-test; re-init) statement Executes loop body as long as test evaluates to TRUE (non-zero). Initialization and re-initialization code included in loop statement. Note: Test is evaluated before executing loop body F init test T loop_body re-init 69 Generating Code for For for (i = 0; i < 10; i++) prf( %d, i); LOOP This is the same as the while example! DONE ; init AND R0, R0, #0 STR R0, R6, #0 ; i = 0 ; test LDR R0, R6, #0 ; load i ADD R0, R0, #-10 BRzp DONE ; loop body LDR R0, R6, #0 ; load i... <prf>... ; re-init ADD R0, R0, #1 ; incr i STR R0, R6, #0 BR LOOP ; test again ; next statement 70 Example For Loops /* -- what is the output of this loop? -- */ for (i = 0; i <= 10; i++) prf("%d ", i); /* -- what does this one output? -- */ letter = 'a'; for (c = 0; c < 26; c++) prf("%c ", letter+c); /* -- what does this loop do? -- */ numberofones = 0; for (bitnum = 0; bitnum < 16; bitnum++) if (inputvalue & (1 << bitnum)) numberofones++; Nested Loops Loop body can (of course) be another loop /* pr a multiplication table */ for (mp1 = 0; mp1 < 10; mp1++) for (mp2 = 0; mp2 < 10; mp2++) prf( %d\t, mp1*mp2); prf( \n ); 71 72

19 Another Nested Loop Here, test for the inner loop depends on counter variable of outer loop for (outer = 1; outer <= input; outer++) for (inner = 0; inner < outer; inner++) sum += inner; For vs. While In general: For loop is preferred for counter-based loops Explicit counter variable Easy to see how counter is modified each loop While loop is preferred for sentinel-based loops Test checks for sentinel value. Either kind of loop can be expressed as other, so really a matter of style and readability Do-While Break and Continue do loop_body; while (test); T loop_body test break; used only in switch statement or iteration statement passes control out of the nearest (loop or switch) statement containing it to the statement immediately following usually used to exit a loop before terminating condition occurs (or to exit switch statement when case is done) F continue; Executes loop body as long as test evaluates to TRUE (non-zero). Note: Test is evaluated after executing loop body used only in iteration statement terminates the execution of the loop body for this iteration loop expression is evaluated to see whether another iteration should be performed if for loop, also executes the re-initializer 75 76

20 Example What does the following loop do? for (i = 0; i <= 20; i++) if (i%2 == 0) continue; prf("%d ", i); What would be an easier way to write this? Switch switch (expression) case const1: action1; break; case const2: action2; break; default: action3; evaluate expression = const1? F = const2? F action3 T T action1 action2 What happens if break instead of continue? 77 Alternative to long if-else chain. If break is not used, then case "falls through" to the next. 78 Switch Example /* same as month example for if-else */ switch (month) case 4: case 6: case 9: case 11: prf( Month has 30 days.\n ); break; case 1: case 3: /* some cases omitted for brevity...*/ prf( Month has 31 days.\n ); break; case 2: prf( Month has 28 or 29 days.\n ); break; default: prf( Don t know that month.\n ); More About Switch Case expressions must be constant case i: /* illegal if i is a variable */ If no break, then next case is also executed switch (a) case 1: prf( A ); case 2: prf( B ); default: prf( C ); If a is 1, prs ABC. If a is 2, prs BC. Otherwise, prs C

21 Enumerations Keyword enum declares a new type enum colors RED, GREEN, BLUE, GREEN, YELLOW, MAUVE ; RED is now 0, GREEN is 1, etc. Gives meaning to constants, groups constants enum colors house_color; house_color = get_color(); switch (house_color) case RED: /* code here */ break; /* more here */ Enums are just s, but can provide more type checking Warning on assignment (example: house_color = 85;) Warning on partial switch statement C++ adds even more checking support 81 Example: Searching for Substring Have user type in a line of text (ending with linefeed) and pr the number of occurrences of "the" Reading characters one at a time Use the getchar() function -- returns a single character Don't need to store input string; look for substring as characters are being typed Similar to state machine: based on characters seen, move toward success state or move back to start state Switch statement is a good match to state machine 82 Substring: State machine to flow chart other 't' 't' no match 't' matched 't' 'h' matched 'th' 'e' other other increment count read char match = 0 F match = 1 F match = 2 F T T T if 't', match=1 if 'h', match=2 if 't', match=1 else match=0 if 'e', count++ and match = 0 if 't', match=1 else match=0 83 Substring: Code (Part 1) #include <stdio.h> enum state NO_MATCH, ONE_MATCH, TWO_MATCHES ; main() char key; /* input character from user */ match = NO_MATCH ; /* state of matching */ count = 0; /* number of substring matches */ /* Read character until newline is typed */ key = getchar(); while (key!= '\n') /* Action depends on number of matches so far */ switch (match) See next two slides for contents of switch statement key = getchar(); prf("number of matches = %d\n", count); 84

22 Substring: Code (Part 2) Substring: Code (Part 3) case NO_MATCH: /* starting - no matches yet */ if (key == 't') match = ONE_MATCH; else match = NO_MATCH; break; case ONE_MATCH: /* 't' has been matched */ if (key == 'h') match = TWO_MATCHES; else if (key == 't') match = ONE_MATCH; else match = NO_MATCH; break; case TWO_MATCHES: /* 'th' has been matched */ if (key == 'e') count++; /* increment count */ match = NO_MATCH; /* go to starting po */ else if (key == 't') match = ONE_MATCH; else match = NO_MATCH; break; C and the Right Shift Operator (>>) Does right shift sign extend or not? Answer: Yes and No Unsigned values: zero extend unsigned x = ~0; Then, (x >> 10) will have 10 leading zeros Signed values: Right shifting a signed quantity will fill with sign bits ( arithmetic shift ) on some machines and with 0-bits ( logical shift ) on others. - Kernighan and Ritchie In practice, it does sign extend! x = ~0; /* signed */!Then, (x >> 10) will still be all 1s Chapter 14 Functions Based on slides McGraw-Hill Additional material 2004/2005/2006 Lewis/Martin 87

23 Function Smaller, simpler, subcomponent of program Provides abstraction Hide low-level details Give high-level structure to program, easier to understand overall program flow Enables separable, independent development C functions Zero or multiple arguments (or parameters) passed in Single result returned (optional) Return value is always a particular type In other languages, called procedures, subroutines,... Example of High-Level Structure void main() setup_board(); /* place pieces on board */ determine_sides(); /* choose black/white */ /* Play game */ while (no_outcome_yet()) whites_turn(); blacks_turn(); Structure of program is evident, even without knowing implementation Functions in C Definition type of return value factorial( n) i; result = 1; for (i = 1; i <= n; i++) result = result * i; return result; Function call -- used in expression a = x + factorial(f + g); name of function types of all arguments 1. evaluate arguments 2. execute function 3. use return value in expression exits function with specified return value 91 Implementing Functions and Variables in LC-3 We ve talked about Variables!Local!Global Functions!Parameter passing!return values What does the assembly code look like for these idioms? Important notes Different compilers for different ISAs do things differently As long as a compiler is consistent We re straying from the book s version to simplify things!leaving out the R5 frame poer 92

24 Allocating Space for Variables Local Variable Storage Global data section All global variables stored here (actually all static variables) R4 pos to beginning Run-time stack Used for local variables R6 pos to top of stack New frame for each block (goes away when block exited) Offset = distance from beginning of storage area Global: LDR R1, R4, #4 0x0000 instructions global data run-time stack PC R4 R6 Local variables stored in activation record (stack frame) Symbol table offset gives the distance from the base of the frame A new frame is pushed on the run-time stack each time block is entered R6 is the stack poer holds address of current top of run-time stack Because stack grows downward, stack poer is the smallest address of the frame, and variable offsets are >= 0. R6 amount hours minute seconds Local: LDR R2, R6, #3 0xFFFF Symbol Table Compiler tracks each symbol (identifiers) and its location In assembler, all identifiers were labels In compiler, identifiers are variables Compiler keeps more information Name (identifier) Type Location in memory Scope Name Type Offset Scope amount hours minutes seconds double main main main main Symbol Table Example main() seconds; minutes; hours; double amount; R6 R6 Name Type Offset Scope amount hours minutes seconds amount hours minute seconds double main main main main 95 96

25 Example: Compiling to LC-3 Example: Code Generation #include <stdio.h> inglobal; main() inlocal; outlocala; outlocalb; Name inglobal inlocal outlocala Type Offset global main main Scope ; main ; inlocal = 5 AND R0, R0, #0 ADD R0, R0, #5 ; inlocal = 5 STR R0, R6, #2 ; (offset = 2) /* initialize */ inlocal = 5; inglobal = 3; outlocalb /* perform calculations */ outlocala = inlocal & ~inglobal; outlocalb = (inlocal + inglobal) + outlocalb; main /* pr results */ prf("the results are: outlocala = %d, outlocalb = %d\n", outlocala, outlocalb); 0 ; inglobal = 3 AND R0, R0, #0 ADD R0, R0, #3 ; inglobal = 3 STR R0, R4, #0 ; (offset = 0) Example (continued) ; first statement: ; outlocala = inlocal & ~inglobal; Example (continued) R0 ;outlocalb = (inlocal + inglobal) + outlocala; LDR R0, R6, #2 ; get inlocal(offset = 2) LDR R1, R4, #0 ; get inglobal NOT R1, R1 ; ~inglobal AND R2, R0, R1 ; inlocal & ~inglobal STR R2, R6, #1 ; store in outlocala ; (offset = 1) LDR R0, R6, #2 ; inlocal LDR R1, R4, #0 ; inglobal ADD R0, R0, R1 ; R0 is sum LDR R1, R6, #1 ; outlocala ADD R2, R0, R1 ; R2 is sum STR R2, R6, #0 ; outlocalb (offset = 0)

26 Implementing Functions Activation record Information about each function, including arguments and local variables Also stored on run-time stack Calling function copy args o stack or regs call function get result Called function allocate activation record save registers execute code put result in AR or reg pop activation record return Run-Time Stack for Functions Memory Memory Memory R6 func R6 main main main Before call During call After call R Activation Record func( a, b) w, x, y;... return y; Name Type Offset Scope R6 bookkeeping y x w save R0 return addr. (R7) return value a b locals args Activation Record Bookkeeping Return value Space for value returned by function Allocated even if function does not return a value Return address Save poer to next instruction in calling function Convenient location to store R7!in case another function (JSR) is called b a ret. value w x y func func func func func func Save registers Save all other registers used (but not R6, and often not R4)

27 Function Call Example main() x, y, val; x = 10; y = 11; val = max(x + 10, y); return val; max( a, b) result; result = a; if (b > a) result = b; return result; main s view R result save R0 save R1 save R7 max return value a b val y x main return value R6 max s view Main Function (1 of 2) MAIN ADD R6, R6, #-4 ; allocate frame AND R0, R0, #0 ; x = 10 ADD R0, R0, #10 STR R0, R6, #2 AND R0, R0, #0 ; y = 11 ADD R0, R0, #11 STR R0, R6, #1 LDR R0, R6, #1 STR R0, R6, #-1 ; load y o R0 ; 2nd argument LDR R1, R6, #2 ; load x o R1 ADD R1, R1, #10 ; R1 = x + 10 STR R1, R6, #-2 ; 1st argument JSR MAX ; more here ; call max function Max Function MAX ADD R6, R6, #-7 ; allocate frame STR R7, R6, #3 ; save R7 (link register) STR R1, R6, #2 ; save R1 STR R0, R6, #1 ; save R0 LDR R0, R6, #5 STR R0, R6, #0 ; load "a" ; store "a" o "result" LDR R1, R6, #6 ; load "b" NOT R1, R1 ; calculate -b ADD R1, R1, 1 LDR R0, R6, #5 ; load "a" ADD R0, R1, R0 ; compare BRp AFTER LDR R0, R6, #6 STR R0, R6, #0 ; load "b" ; store "b" o "result" AFTER LDR R0, R6, #0 ; load "result" STR R0, R6, #4 ; store "result" o return value LDR R0, R6, #1 LDR R1, R6, #2 LDR R7, R6, #3 ADD R6, R6, #7 RET ; restore R0 ; restore R1 ; restore R7 (link register) ; pop stack 107 Main Function (2 of 2) ; previous code here JSR MAX LDR R0, R6, #-3 STR R0, R6, #0 LDR R0, R6, #0 STR R0, R6, #3 ADD R6, R6, #4 RET ; call max function ; read return value of max ; put value o local "val" ; load "val" ; put "val" o main s ; return value ; pop stack 108

28 Summary of LC-3 Function Call Implementation 1. Caller places arguments on stack (last to first) 2. Caller invokes subroutine (JSR) 3. Callee allocates frame 4. Callee saves R7 and other registers 5. Callee executes function code 6. Callee stores result o return value slot 7. Callee restores registers 8. Callee deallocates frame (local vars, other registers) 9. Callee returns (RET or JMP R7) 10. Caller loads return value 11. Caller resumes computation 109 Callee versus Caller Saved Registers Callee saved registers In our examples, the callee saved and restored registers Saves/restores any registers it modifies What if a you wants R7 to be preserved across a call? Before call: caller saves it on the stack After call: caller restores it from the stack Caller saved registers R7 is an example of a caller saved register Value assumed destroyed across calls Only needs to save R7 when it s in use Which is better? Callee or Caller saved registers? Neither: many ISA calling conventions specify some of each 110 Compilers are Smart(er) In our examples, variables always stored in memory Read from stack, written to stack Compiler will perform code optimizations Keeps many variables in registers Avoids many save/restores of registers Why? Passing parameter values in registers First few parameters in registers Return value in register Like in your homework projects Again, why? 111